Do European Banks with a Covered Bond Program still

issue Asset-Backed Securities for funding?∗

Nils Boesel † Clemens Kool ‡ Stefano Lugo §

January 13, 2016

Abstract

The decline in the issuance of Asset-Backed Securities (ABS) since the financial crisis and

the comparative advantage of Covered Bonds (CBs) as a funding alternative to ABS raise

the question whether banks still issue ABS as a mean to receive funding. Employing double-

hurdle regression models on a dataset of 134 European banks observed during the period from

2007 to 2013, this study reveals that banks with a Covered Bond Program (CBP) securitize

ceteris paribus less of their assets. The estimated difference in ABS issuance is mainly driven by

banks more likely to issue ABS as a funding tool, rather than trying to manage their credit risk

exposure or meeting regulatory capital requirements. Consistently, a worse liquidity/funding

position results in higher levels of securitization only for banks without a CBP.

EFM-classification: 510, 520

JEL-classification: G21, G28

Keywords: Securitization, asset-backed securities, covered bonds, bank funding, capital relief

∗Part of this research was performed while Boesel was at the European Central Bank (ECB). We are grateful to

the ECB and J.P.Morgan for granting us access to part of the data used in this study. The views expressed in this

article are the authors’ only, and do not necessarily represent those of the ECB, the Bundesbank, or the CPB.†Deutsche Bundesbank. Wilhelm-Epstein-Strasse 14, 60431 Frankfurt am Main (Germany). Phone: +49 (0) 69

9566 3414. E-mail: nils.boesel@bundesbank.de‡Centraal PlanBureau (CPB) and Utrecht University School of Economics (U.S.E.). Kriekenpitplein 21-22, 3584

EC Utrecht (Netherlands). Phone: +31 (0)30 253 9813. E-mail: c.j.m.kool@uu.nl§Utrecht University School of Economics (U.S.E.). Kriekenpitplein 21-22, 3584 EC Utrecht (Netherlands). Phone:

+31(0)30 253 7297. E-mail: s.lugo@uu.nl (contact author).

1

1 Introduction

The issuance of Asset-Backed Securities (ABS)1 in Europe has decreased dramatically in size since

the advent of the 2007 financial crisis. According to SIFMA/AFME data,2 the amount of Euro

denominated structured finance products (excluding retained securities) issued in 2007 was Euro

418 billion; by 2012, the figure had shrunk to Euro 253 billion. Clearly, one of the main driver of this

trend is the central role that structured finance products have played in the unfolding of the crisis

in the United States (e.g., Coval, Jurek, and Stafford, 2009; Longstaff, 2010; Gorton and Metrick,

2012). European financial institutions were among the main investors in US ABS before the advent

of the crisis (Bertaut, DeMarco, Kamin, and Tryon, 2012), creating a direct link for contagion. This

has raised investors skepticism toward this class of complex financial products (e.g., Celerier and

Vallee, 2014). Despite the potential conflicts of interest and misaligned incentives that characterize

the originate-to-distribute banking scheme, European official institutions still regard ABS as key

instruments to allow banks to raise the necessary capital to provide consumers and Small and

Medium Enterprises (SMEs) with credit (European Commission, 2015). Thus, it is important to

understand what currently drives European banks decision to securitize their assets.

While ABS issuance has shrunk by around half between 2007 and 2012, during the same period

the issuance of Euro-denominated Covered Bonds (CBs) has actually increased, from Euro 332

billion to 405 billion, according to European Covered Bond Council (ECBC) data.3 Van Rixtel and

Gasperini (2013) report that the share of CBs over total gross bonds issued by European banks has

increased from 26% in 2007 to 42% in 2012. Some commentators (e.g., European Central Bank,

2011) have argued that banks have replaced ABS with CBs issuance since the advent of the crisis.

To a certain extent ABS and CBs can in fact be considered as substitutes. Both are fixed-income

securities backed by a ring-fenced pool of collateral assets. However, ABS and CBs presents also

some key differences, as discussed at length in Section 2. Most notably, the transfer of credit risk

from originators to investors is less complete with CBs than with ABSwhile CBs provide investors

via a dual recourse mechanism with an extra layer of protection. This has two direct consequences.

On the one hand, the higher investor protection entailed by CBs results in lower yields, and thus

lower costs of funding. On the other hand, CBs are not an effective way for the originator to reduce

its credit risk exposure (Packer, Stever, and Upper, 2007). This and other material differences

between ABS and CBs can provide the latter with a comparative advantage as a funding tool,

albeit not as a way to reduce credit risk exposure and/or Risk-Weighted Assets (RWA). Not every

1We use throughout this paper the term ABS to generally identify structured finance products resulting from

the securitization of extant financial assets via a Special Purpose Vehicle (SPV). The term thus encompasses also

Mortgage-Backed Securities (MBS), Collateralized Debt Obligations (CDO), and Collateralized Loan Obligations

(CLO).2Data accessed at www.sifma.org on November 11th, 2015.3Data accessed at http://ecbc.hypo.org on November 24th, 2015.

2

bank is able to issue CBs however, as Covered Bond Programs (CBPs) are strictly regulated in

most European countries. On the contrary, contractually-based ABS allow for more flexibility and

can be issued as a one-off transaction. Thus, securitization can still be a viable source of liquidity

for banks with no access to the CBs market.

In this study, we address how the ability to issue CBs as an alternative to ABS affects the issuance

activity of structured finance products by European banks. We use a sample of 775 observations for

134 European banks observed between 2007 and 2013, and investigate how having a CBP affects

how much and why banks issue ABS. We do so by employing double-hurdle models, which allow

to account for potential differences in the effect of a bank’s characteristics on if and how much it

engages in securitization.

Our main findings are readily summarized. We show that, all else equal, banks with a CBP are

expected to securitize a lower share of their assets. The difference is around 0.11–0.13 percentage

points, depending on the model. As a term of comparison, banks in our sample securitize on average

0.18% of their assets. Consistent with the comparative advantage of CBs as a funding tool, the

difference between the level of ABS issuance for banks with and without a CBP is mainly driven

by banks with a low level of assets liquidity. Proxies for the level of a bank’s credit risk exposure

or need to meet regulatory capital requirements do not appear to play an equally relevant role.

Coherently, we find a negative marginal effect of asset’s liquidity on securitization levels for banks

without a CBP, but not for banks able to issue CBs. The difference between the two marginal

effects is statistically significant, and increases with the level of asset’s illiquidity. No significant

difference in the marginal effect of proxies for a bank’s assets credit risk and regulatory capital

position is found across banks with and without a CBP.

This paper contributes to a fast growing literature addressing banks securitization activities

(Maddaloni and Peydro, 2011; Pagano and Volpin, 2012; Jiang, Nelson, and Vytlacil, 2013; Wang

and Xia, 2014, among others), and in particular to the literature focusing on how individual

characteristics of European banks can explain their ABS issuance levels (Affinito and Tagliaferri,

2010; Cardone-Riportella, Samaniego-Medina, and Trujillo-Ponce, 2010; Farruggio and Uhde, 2015).

These studies have generally identified a bank’s need for liquidity as one of the main determinant

for securitization. However, they typically do not consider the role played by the availability of

alternative sources of liquidity. To the best of our knowledge, this is the first study to show how

the ability to issue CBs influences a bank’s securitization activity–and particularly so during the

recent crisis period. In a related study, Carbo-Valverde, Fernandez, and Rosen (2011) examine the

determinants of both CBs and Mortgage-Backed Securities (MBS) issuance as independent from

each other. They find that banks issue CBs, but not MBS, to meet their liquidity needs. By looking

at a bank’s ability to issue CBs and focusing on the issuance of structured finance products, we com-

plement their results by showing that European banks still issue ABS for funding/liquidity reasons,

but only when they cannot resort to CBs as a viable alternative funding tool. This is consistent

3

with the idea that–when available–covered bonds currently constitute a preferred funding tool over

Asset-Backed Securities.

The rest of this paper is organized as follows. Section 2 reviews the rationales for securitization in

light of the existing literature and discusses the material differences between Asset-Backed Securities

and Covered Bonds. Section 3 presents the data and methodology used in this study. Empirical

results are presented and discussed in Section 4. Section 5 concludes with some final remarks.

2 Background and Empirical Predictions

In this Section we review the extant literature on banks rationales for securitization (Section 2.1)

and discuss the main differences between ABS and CBs culminating in our empirical predictions

(Section 2.2).

2.1 Rationales for Securitization

Extant literature identifies three main rationales for securitization relating to individual character-

istics of the issuer.

The first reason for banks to issue structured finance products is to satisfy their funding/liquidity

needs. Securitization allows to efficiently sell illiquid assets such as mortgages and corporate loans

and transform them into new liquidity to be retained or invested. Securitization as a form of raising

funds can often be preferable to raising new retail deposits. The latter can be more costly, for

example because of higher capital requirements; they are also redeemable at any point in time,

leaving the bank exposed to bank runs. ABS issuance can instead be seen as a form of medium

to long-term financing. Several previous studies (e.g., Affinito and Tagliaferri, 2010; Cardone-

Riportella, Samaniego-Medina, and Trujillo-Ponce, 2010; Farruggio and Uhde, 2015) have identified

the level of a bank’s assets liquidity as one of the main driver of securitization.

The second reason for securitization relates to the transfer of credit risk to investors. By selling

the underlying assets to a Special Purpose Vehicle (SPV), the bank removes risky assets from

its balance sheet and passes the credit risk to the investors. The credit risk of a bank’s asset is

thus expected to play a relevant role in the determination of its securitization activity. However,

the direction of such effect is not trivial. On the one hand, banks with more risky assets can

have a greater incentive toward securitization. On the other hand, given the existing information

asymmetries only banks with a high asset quality and a good reputation might be able to securitize

their assets without incurring a considerable discount on their face value (e.g., Albertazzi, Eramo,

Gambacorta, and Salleo, 2015; Ambrose, LaCour-Little, and Sanders, 2005; An, Deng, and Gabriel,

2011; Calomiris and Mason, 2004). Moreover, banks often have to retain the equity (i.e., the first

loss) tranche of the ABS on their balance sheet (DeMarzo, 2005). As a result of these contrasting

effects, previous empirical studies have found support for both a positive (Affinito and Tagliaferri,

4

2010; Bannier and Hansel, 2006) and a negative (Calem and LaCour-Little, 2004; Farruggio and

Uhde, 2015) relationship between asset’s credit risk and level of securitization.

The third rationale for securitization is linked to a bank’s need to meet its regulatory capital re-

quirements. By transferring credit risk out of its balance sheet, a bank can reduce its Risk-Weighted

Assets (RWA) and thus its need for Tier I and II capital as required by Basel rules. Securitization

can even allow a bank to engage in regulatory capital arbitrage, i.e., reducing capital requirement

while not effectively reducing credit risk exposure. This possibility was particularly evident under

Basel I rules, as illustrated by Jones (2000). While Basel II tries to significantly mitigate the scope

for regulatory capital arbitrage (Blum, 2008; Fabozzi, Davis, and Choudhry, 2007), such opportu-

nities can still be in place to a lesser extent with the current regulatory framework (e.g., Calem

and LaCour-Little, 2004). There is some empirical evidence that regulatory capital requirements

might be a significant driver for securitization, especially during the Basel I regime period (Acharya,

Schnabl, and Suarez, 2013; Affinito and Tagliaferri, 2010; Ambrose, LaCour-Little, and Sanders,

2005).

2.2 Asset-Backed Securities versus Covered Bonds

As already stated in Section 1, ABS and CBs can at first glance be considered as substitute. Both

are in fact fixed-income securities backed by a determined pool of collateral which is ring-fenced

from other liabilities of the issuer. There are however notable differences making them very different

kind of instruments both from the point of view of issuers and investors.

2.2.1 Asset-Backed Securities

ABS are technically issued by an ad-hoc created Special Purpose Vehicle (SPV). The SPV buys the

collateral from the depositor (the entity collecting the underlying assets and creating the SPV) and

issues structured finance products backed by such collateral. Proceeds of the sale of ABS to investors

are ultimately used to pay the depositor. The latter is typically also the originator and servicer

of the assets (e.g., the bank granting a firm with the loan subsequently securitized and collecting

the regular interest payments), but assets can also be originated by third parties. Securitization is

mostly based on contractual mechanisms, thus allowing for a high degree of flexibility (European

Banking Authority, 2014a). ABS are usually structured in different tranches, allowing for different

levels of exposure to credit, prepayment, and interest rate risk and thereby tailored to varying

investor risk/return appetites. Further credit enhancement, such as subordination, excess spread

and third-party insurance, can also be embedded in the deal.

One of the main consequence of the way ABS are issued is that the SPV provides bankruptcy

remoteness (Ayotte and Gaon, 2011) to the depositor, i.e., investors in ABS only have a claim on the

assets effectively owned by the SPV. Even if Basel II requires originators to retain a significant net

5

economic interest in the underlying assets (Basurto, Jones, Lindner, and Blankenheim, 2015), by

design the structured finance market is thus exposed to agency problems (see for example Ashcraft

and Schuermann, 2006). Originators have superior knowledge about the true quality of assets. This

can lead to adverse selection, as already discussed in Section 2.1. At the same time, bankruptcy

remoteness entails that, by issuing ABS, banks can effectively remove risky assets from their balance

sheet and reduce their RWA computed for regulatory purposes.

2.2.2 Covered Bonds

Differently from ABS, Covered Bonds are issued directly by the bank owning the assets, which

remains ultimately liable for the payments due to investors. The issuer designates a pool of assets

to be pledged as collateral, but retains iton its balance sheet. In case assets used as collateral no

longer meet the quality criteria mandated by national laws or if prepayments occur, the issuer has

to increase/replace the assets in the pool. Should the bank issuing covered bonds face financial

distress, the ring-fenced assets are shielded from claims of other creditors of the bank. If, in case

of default, the collateral is not sufficient to cover the claims of CB investors, CBs entail the same

seniority over the remaining assets of the bank as unsecured senior bonds. As such, the collateral

pool for a CB has to be considered more as a form of credit enhancement, rather than the effective

source of cash flows paid to investors. Moreover, CBs are additionally argued to benefit from an

implicit state guarantee that does not apply to ABS. Volk and Will (2012); European Banking

Authority (2014b) stress that, contrary to ABS, a strong tendency for government support for CBs

through support for the issuing bank could be observed in various recent cases. This is partially

the result of CBs underlying assets being retained on a bank’s balance sheet, rather than being sold

to an SPV. As a result, CBs tend to be characterized by significantly lower yields than other debt

securities issued by the same bank (Packer, Stever, and Upper, 2007), and the spread between CBs

and government bonds is often considered to merely reflect a liquidity premium (e.g., Kempf, Korn,

and Uhrig-Homburg, 2012).

An other key difference between ABS and CBs is that the quality of eligible assets to be used as

collateral for a CB is strictly regulated by national legislation and needs to be regularly reassessed

by national agencies. At European level, the Capital Requirement Directive (CRD) only recognize

as CBs securities issued under such legislations (for a brief review of CBP legislative frameworks,

see for example Packer, Stever, and Upper, 2007). Notably, many European national regulations

also dictate the required characteristics for an institution to be authorized to have a Covered Bond

Program (CBP) and thus to issue CBs. Thereby, in some jurisdictions the prospective issuer is

required to apply for a covered bond specific licence, while in jurisdictions without a specific licence,

detailed requirements are laid out concerning both issuer and cover pool characteristics European

Banking Authority (2014b). For example, the law governing German Pfanbriefgesetz requires banks

to prove their intention to engage in CBs issuance on a regular basis; a license to issue CBs can

6

be revoked if a bank does not issue CBs for two years. Under the French legislative framework the

issuing entity is even limited in the range of business activities it conducts. In sum, while European

regulatory frameworks allow for considerable flexibility to issue ABS, they are more stringent on

the key characteristics of both the CB issuer as well as the underlying asset pool.

2.2.3 Empirical predictions

From the discussion above on the different characteristics of ABS and CBs, summarized in Table 1,

it follows that CBs have a comparative advantage over ABS in that they are less affected by agency

problems. This is all the more relevant during a crisis period, when investor confidence is eroded.

We thus expect that, all else equal, during the crisis period investigated in this study banks allowed

to issue CBs securitize less than banks without a CBP.

[Insert Table 1 about here]

In this context, CBs can be regarded as a cheaper funding tool than ABS. Indeed, Carbo-

Valverde, Fernandez, and Rosen (2011) show that banks are more likely to issue CBs (but not

ABS) for liquidity reasons. Consistently, we expect the securitization activity of a bank with a

CBP to be independent from its funding needs. However, for banks not able to resort to CBs, ABS

can still constitute a relevant funding tool. All in all, the role of liquidity in explaining the intensity

of ABS issuance is predicted to differ significantly between banks with and without a CBP.

Whereas CBs may constitute a preferable substitute for ABS as a funding tool, this is not the

case for the other two main rationales for securitization illustrated in Section 2.1, i.e., transfering

credit risk exposure and meeting regulatory capital requirements. This is because the cover pool

is retained on a bank’s balance sheet, and credit risk is not transferred to the investors. Covered

bonds thus do not allow the issuer to pursue these goals the same way that ABS do. As such, the

ability to issue CBs should not influence the role played by these two rationales in explaining the

observed levels of securitization.

3 Data and Methodology

This section presents the dataset (Section 3.1) and the methodology (Section 3.2) used in our

analyses.

3.1 Data

The starting point in building our sample is a dataset retrieved from Bankscope. We consider

securitizing as well as non-securitizing banks across the EU-13 countries in the European Monetary

7

Union (EMU).4 The sample covers the period from 2007 to 2013. We focus on this period for

two reasons. First, we are particularly interested in the securitization activity of banks since the

2007 crisis eroded investors confidence in ABS. Second, the implementation of Basel II capital

requirement started in the considered countries on January 1st, 2007 (ECB, 2007). Securitization

activities before and after the passage from Basel I to Basel II might not be comparable, as the

two regulations differ significantly in the way securitization is accounted for (Fabozzi, Davis, and

Choudhry, 2007). The Bankscope dataset initially includes 179 banks from the considered countries.

After excluding non-independent banks (i.e., banks whose ultimate owner is another bank), and

bank-year observations where the necessary bank-level variables–described below–are missing, we

are left with 775 observations for 134 banks. We refer henceforth to these as the dataset used in

this study. Table 2 presents the distribution of the dataset by country and year.

[Insert Table 2 about here]

3.1.1 Asset-Backed Securities issuance

Our dependent variables aim at capturing if and how much European banks resort to securitization.

Data on the issuance of structured finance products are provided by J.P. Morgan and retrieved from

their “International ABS & CB Research” database, which records amongst others the involved

counterparties, the tranche size and currency as well as the collateral country for all ABS issuances

priced worldwide. . The dataset initially includes 7,174 Euro-denominated tranches issued between

2007 and 2013, for a total value of Euro 2.25 trillion. As a term of comparison, the total amount of

European ABSs, Collateralized Debt Obligations (CDOs) and Mortgage-Backed Securities (MBSs)

issued over the same period amounts to Euro 2.64 trillion, according to SIFMA/AFME data.5

We carefully hand-match the structured finance dataset with the 775 bank-year observations in our

sample, in order to make sure we identify all securities originated by the 134 banks considered in this

study. Of the 7,174 securities in the J.P. Morgan dataset, 2,037 are originated by one of the banks

in our sample. As it is customary in the literature (e.g., Affinito and Tagliaferri, 2010; Farruggio

and Uhde, 2015), for each bank i in year t we measure the level of securitization as the total value of

structured finance products issued by bank i in year t, scaled by the bank’s consolidated total assets

as reported in Bankscope (SEC ratioi,t). SECi,t is an indicator equal to 1 if SEC ratioi,t is strictly

positive (i.e., if bank i issued at least one security in year t) and 0 otherwise. Following Affinito

4The countries initially considered are thus Austria, Belgium, Germany, Spain, Finland, France, Greece, Ireland,

Italy, Luxembourg, Netherlands, and Portugal. We later exclude Luxembourg because of no usable observations.

From the EU-13 countries, we exclude the UK for two reasons. First, we want to focus on banks within the EMU

area. Second, several financial institutions in the UK issue unregulated, “structured” covered bonds without having

a CBP. This is generally not the case in the countries considered. See www.ecbc.eu/framework/list5Data accessed at www.sifma.org on October 19th, 2015.

8

and Tagliaferri (2010), we exclude from the computation of SEC ratio and SEC those securities

originated to be retained. Of the 134 banks in the sample, 36 have issued (and not retained) at

least one structured finance product in the period considered (74 bank-year observations). Table 3

reports some descriptive statistics for the securitization activity of these banks.

[Insert Table 3 about here]

On average securitizing banks create around 7 deals (17 tranches) per year, for a total yearly

amount of Euro 2.4 billion. Securitization directly-backed by residential mortgages (RMBS) are

by far the most common, representing around half of the securitized amount; securities backed by

loans to Small and Medium Enterprises (SMEs) follow at around 14%.

3.1.2 Covered Bond program

Our main explanatory variable of interest is an indicator (CBP ) capturing whether a bank has a

Covered Bond Program or not. In order to build CBP we start from the technical reports of the

Norddeutsche Landesbank (2013), which include a list of banks issuing Covered Bonds (CBs). If a

bank is in the list of CBs issuer, then CBP is set equal to 1 for that bank; if not, then the indicator is

set equal to 0. We further check whether the 134 banks are included in the list of banks with a CBP

compiled by the European Covered Bond Council (ECBC).6 Finally, we search on Bloomberg each

of the 134 banks considered, to make sure that no bank where CBP = 0 has actually issued CBs,

and viceversa.7 By construction, CBP is a time-invariant characteristic of each bank. Differently

from Carbo-Valverde, Fernandez, and Rosen (2011), who focus on the different rationales for banks

to issue CBs and ABSs, in this study we are interested in how the ability of a bank to issue CBs

affects its securitization activity, which is why we focus on CBP rather than on the CBs issuance

activity in each year. Of the 134 banks in the dataset 70 appear to be able to issue covered bonds,

corresponding to 424 observations (55% of the sample).

.

3.1.3 Bank characteristics and other controls

To proxy for banks rationales for securitization, as well as to control for extra variables likely to

influence the phenomena under study, we consider a number of relevant bank’s characteristics. All

of the bank variables described below are computed based on Bankscope data, and are lagged 1

year as it is customary to reduce endogeneity. Whenever possible, we closely follow the definitions

proposed by Farruggio and Uhde (2015), who also investigate the rationales for the securitization

activity of European banks using Bankscope data.

6Data available at http://www.ecbc.eu/issuers7For no bank the value of CBP has changed as a result of these further checks.

9

To proxy for a bank liquidity/funding needs, we use a variable defined as one minus the ra-

tio of net loans to total assets (Liquidity ratio). Observations characterized by lower values of

Liquidity ratio identify banks more likely to securitize for liquidity/funding needs. The ratio of

loan loss reserves to gross loans (Loss to loans) proxies for a bank’s exposure to credit risk. As

discussed in Section 2.1, the credit risk of a bank’s assets can either exhibit a positive or negative

correlation with the intensity of that bank’s securitization activity. To proxy for a bank’s need to

improve its regulatory capital position, we use the ratio of a bank’s Tier 1 capital over its risk-

weighted assets (Tier 1 ratio). We control for a bank’s operative performances using the ratio of

operating profit to equity (ROE). Liquidity ratio, Loss to loans, Tier 1 ratio, and ROE are all

expressed in percentage points. We proxy for a bank’s size using the natural logarithm of its total

assets, expressed in Euro millions (LNTA). Cardone-Riportella, Samaniego-Medina, and Trujillo-

Ponce (2010) show that banks bigger and with better operative performances tend to securitize

more. Finally, we control for a bank’s typology using an indicator equal to 1 if it is a commercial

bank and 0 otherwise (Commercial).

We also include in our analysis two country-level control variables (source: Datastream): the

difference between the 10-year and 3-year yields on government bonds (Slope) and the annual GDP

growth (GDP ). These variables are typically included in studies on securitization to control for the

current status and future prospects of the economy. Banks in growing economies have been found to

ceteris paribus engage more heavily in securitization activities (e.g., Maddaloni and Peydro, 2011).

Table 4 reports descriptive statistics for all of the variables included in this study.

[Insert Table 4 about here]

On average banks in our sample securitize (excluding retained securities) 0.18% of their assets.

This number is slightly lower than reported by Farruggio and Uhde (2015) –consistent with the

inclusion of three more crisis years in our sample– but comparable in magnitude. The share is

1.84% when only observations where SEC = 1 are considered.

3.2 Methodology

Since by definition the issuance of structured finance product cannot be negative, previous studies

(e.g., Farruggio and Uhde, 2015) have often investigated the determinants of securitization by using

a Tobit model. However, such model ignores the two-stage nature of the decision, as banks have

to decide both if and how much of its assets to securitize. The two decisions could in principle be

influenced by different determinants, or by the same determinants in a different way; a traditional

Tobit model does not allow for such flexibility. In this study we thus focus on double-hurdle models,

which allow to tease out the distinction between the decision of if and how much to issue (for a

detailed discussion on this class of models, see Wooldridge, 2010).

10

The first model we consider is a Cragg (1971) truncated normal hurdle model. The model

uses a probit specification for the selection equation, i.e., Pr(SEC = 1 | X) = Φ(Xγ) , and a

truncated (at zero) normal distribution for the amount equation, i.e., f(SEC ratio | X,SEC =

1) = [Φ (Xβ/σ)]−1 φ [(SEC ratio−Xβ) /σ] /σ . X is a vector of explanatory variables, which in

this case we assume to be the same for both the selection and the amount equation.8

The unconditional9 expected value of SEC ratio is then given by

E(SEC ratio | X) = Φ(Xγ) [Xβ + σλ(Xβ/σ)] (1)

where λ(c) is the Inverse Mills Ratio (IMR), i.e., λ(c) = φ(c)/Φ(c).

A potential limitation of the Cragg model is that the error terms for the selection and the amount

equations are assumed to be independent. For this, we also use in our analysis a Tobit II model,

which allows the two terms to be correlated. In this case, we include among the determinants of the

binary decision to issue ABS a dummy variable equal to one if the bank issued ABS securities in the

previous year and zero otherwise (SECt−1). This is done in order to avoid excessive reliance on the

non-linearity of the IMR for identification. The rationale for using SECt−1 for this purpose is that

banks that issued ABS in the recent past are expected to be ceteris paribus more prepared to do so

also in the future; however, the effective amount of assets securitized is more likely dependent on

its contingent needs. The first-time issuance of an ABS, for example, requires a significant build up

of ressources and expertise that can be utilized for future transactions. From an empirical point of

view, SECt−1 appears to be an apt choice for identification. The pairwise correlation between SECt

and SECt−1 is 0.719, statistically significant at the 1% confidence level; the correlation between

SECt−1 and SEC ratio –conditioning on the latter being strictly positive– is instead -0.002, not

statistically significant at customary confidence levels. A likelihood ratio test shows that a Tobit

II model including SECt−1 among the selection determinants significantly (at the 1% confidence

level) outperform a nested model not including it. Including SECt−1 also in the amount equation

does not significantly (at customary confidence levels) increase the likelihood score further.10

With both Cragg and Tobit II models, we address the effect of each explanatory variable x on

the securitization activity of the bank by looking at the Average Marginal Effect (AME) of x on the

unconditional expected value of SEC ratio. As it is customary, for indicators (most notably CBP )

the latter is computed as the difference between the expected unconditional value of SEC ratio

when the variable is equal to one and the expected unconditional value when it is equal to zero.

8Cragg (1971) also proposes an alternative specification where the amount variable is assumed to follow a lognormal

distribution. A Vuong test does not reject the null hypothesis that the two models are equivalent in terms of fitness.

We thus use the truncated normal hurdle model, as it allows for a more straightforward interpretation of the results.9“unconditional” in this setting refers to not conditioning on SEC ratio being strictly positive. The expected

value is of course conditional on the values assumed by the explanatory variables.10In unreported robustness checks, we use for identification an indicator equal to 1 if the bank has issued ABS in

the previous 2 (or 3) years, and zero otherwise. Results are similar to those reported in the paper.

11

4 Results

In this Section we present the empirical results of our study. Section 4.1 focuses on the difference in

the intensity of securitization activities between banks with and without a covered bond program.

Section 4.2 addresses the interaction between the presence of a covered bond program and a bank’s

characteristics in explaining its securitization activity.

4.1 Covered bond program and ABS issuance

Table 5 presents coefficient estimates for a Cragg Model (i) and a Tobit II Model (ii) of the secu-

ritization activity of a bank, excluding retained securities. For each model, we present estimates

for the selection equation (Columns (1) and (4)) and the amount equation (Columns (2) and (5)),

as well as the Average Marginal Effect (AME) on the unconditional expected value of SEC ratio

computed based on those estimates. Each model includes CBP and the control variables illustrated

in Section 3. To account for time trends, we also include year indicators among the explanatory

variables.

[Insert Table 5 about here]

Looking at the results for the Cragg model, we see that the unconditional expected level of

securitization is on average 0.13 percentage points lower when banks have a covered bond program.

The AME is statistically significant at the 5% confidence level. Beside statistical significance the

effect is also economically relevant, as it implies a 67% lower level of securitization relative to the

mean in our sample (0.18%). A very similar result is obtained when a Tobit II model is used.

The AME of CBP is -0.11, statistically significant at the 10% confidence level. Empirical results

reported in Table 5 thus suggest that, as expected, banks with a covered bond program are ceteris

paribus less inclined to issue structured finance products. Before moving to the analysis of how

bank’s characteristics moderate this result, it is worth to briefly discuss the results obtained for the

other explanatory variables.

As predicted by the theory, Liquidity ratio negatively correlates with the intensity of a bank’s

securitization, i.e., more illiquid banks engage more in securitization activities. On average, a 1

percentage point decrease in the level of liquidity of a bank’s asset is associated with a 1-basis

point increase in the unconditional expected share of assets securitized. The effect is statistically

significant at the 1% confidence level with both models. The level of credit risk of a bank’s asset,

as proxied by Loss to loans, is also negatively related to ABS issuance, a result in line with those

reported by previous studies (e.g., Calem and LaCour-Little, 2004; Farruggio and Uhde, 2015).

On average, a 1 percentage point higher level of loss allowances is associated with a reduction

in SEC ratio between 0.03 and 0.07 percentage points, depending on the model. The effect is

12

statistically significant at least at the 10% confidence level. As discussed in Section 2.1, this result

can be explained in terms of asymmetric information between the bank and the investors about the

quality of the collateral. Securitization also decreases with Tier 1 ratio, albeit the average effect is

not statistically significant at customary confidence levels. The direction of the effect is coherent

with the idea that securitization can be used to improve the regulatory capital position of a bank,

and it is consistent with the findings of Affinito and Tagliaferri (2010); Ambrose, LaCour-Little,

and Sanders (2005); Calem and LaCour-Little (2004); Calomiris and Mason (2004); Cebenoyan

and Strahan (2004) among others. Bigger banks (higher LNTA) and banks with better operative

performances (i.e., higher ROE) appear to securitize more, even though the average effect is not

statistically significant for the latter and only statistically significant with Model (i) for the former.

It is interesting to notice how LNTA appears to impact the decision to securitize and the level

of securitization conditional on issuance in opposite directions, confirming the importance of using

double-hurdle models. Finally, both Slope and GDP exhibit a positive AME, a result in line with

those of Maddaloni and Peydro (2011).

4.2 Moderating role of rationales for securitization

In this Section we investigate the moderating role of a bank’s characteristics on the relationship

between CBP and securitization activities. In particular, we focus on the proxies for the three main

rationales for securitization (i.e., funding/liquidity, credit risk, and regulatory capital requirements).

As discussed in Section 2.2, we expect the difference between banks with and without a covered

bond program to be especially prominent for banks more likely to securitize for liquidity reasons.

To better gauge the interaction between rationales for securitization and the availability of covered

bonds as an alternative to ABSs, we estimate three Tobit II models including the cross-product

of CBP and Liquidity ratio (i), Loss to loans (ii), or Tier 1 ratio (iii). Coefficients estimates are

reported in Table 6. The three models include all of the control variables of Model (ii) of Table 5.

[Insert Table 6 about here]

Figure 1 illustrates the average marginal effects computed based on coefficient estimates for

Models (i)–(iii) of Table 6. For each model, we compute the AME of CBP for different values of

bank’s characteristic x, where x is either Liquidity ratio, Loss to loans, or Tier 1 ratio, depending

on the model. We also compute the difference in the average marginal effect of x itself between

banks with and without a covered bond program for different levels of x. We expect the difference

in the unconditional expected level of securitization between banks with and without a covered

bond program to be bigger for more illiquid banks, i.e., for banks more likely to issue ABSs for

liquidity/funding reasons. We also expect a stronger (negative) marginal effect of Liquidity ratio

on the level of securitization for banks without a covered bond program, especially when the bank’s

13

asset are very illiquid (i.e., when Liquidity ratio is low). For the reasons discussed in Section 2.2,

we do not instead expect a significant difference in the ABS issuance of banks with and without a

covered bond program when credit risk and regulatory capital requirements needs are more likely

to be the main determinants of securitization.

[Insert Figure 1 about here]

As expected, the marginal effect of CBP monotonically decreases with the level of Liquidityratio.

The difference in the unconditional expected value of SEC ratio between banks with and without a

covered bond program is statistically significant at the 5% (1%) confidence level when Liquidityratio

is lower than 41% (35%). For extremely illiquid banks –i.e., Liquidity ratio approaching 0– the

availability of a covered bond program is estimated to be associated with around 1-percentage

points lower levels of securitization. The negative effect of Liquidity ratio on the unconditional

expected value of SEC ratio is significantly (at the 5% confidence level) stronger for banks without

a covered bond program throughout most of the variable’s domain,and it is bigger the more illiquid

the bank is. For example, for a bank characterized by a Liquidity ratio of 20%, a 1 percentage point

decrease in asset’s liquidity is associated with a 0.02 percentage point stronger negative effect on

securitization for banks without a covered bond program. When Liquidity ratio is equal to 10%,

the difference raises to 0.03 percentage points. Interestingly, this functional form is entirely driven

by the AME of banks without a covered bond program, as illustrated in Figure 2. Whereas the

AME of Liquidity ratio is higher (in absolute terms) the more banks are illiquid, it is close to zero

and not statistically significant throughout the whole range for banks with a covered bond program.

The latter observation is consistent with the results reported by Carbo-Valverde, Fernandez, and

Rosen (2011), who find that banks do not securitize for funding purposes. We show that this is true

only for banks who actually have access to covered bonds as an alternative funding instrument.

[Insert Figure 2 about here]

As for the other explanatory variables, Figure 1 shows that credit risk does not seem to play a

big role in explaining potential differences in the securitization activity of banks with and without

a covered bond program. The difference in the average unconditional expected value of SEC ratio

between the two is statistically significant at the 5% confidence level only for very low levels (< 5%)

of Loss to loans, i.e., when banks are very unlikely to issue ABSs in order to remove the most risky

assets from their balance sheets. As expected, also regulatory capital does not appear to play a

significant role in explaining differences in the securitization activity of the two groups of banks.

All in all, results presented in this Section suggest that the difference in the securitization activity

of banks with and without a covered bond programs is mainly driven by banks most likely to issue

14

ABS as a funding tool. This is consistent with the idea that CBs currently have a comparative

advantage over ABS as a funding tool, albeit they do not serve as an effective instrument to reduce

a bank’s credit risk exposure.

5 Conclusions

In this paper we investigate how the ability to issue Covered Bonds (CBs) affects the securitization

activity of European banks since the advent of the 2007 financial crisis. Asset-Backed Securities

(ABS) and CBs are both fixed-income securities linked to ring-fenced assets. However, CBs do not

allow banks to transfer credit risk to the same extent that ABS do.

The resulting mitigation of asymmetric information and agency problems provide CBs with a

comparative advantage as a funding tool over ABS, albeit not as a way to transfer credit risk.

Using a sample of 775 observations for 134 European banks between 2007 and 2013, we show

that banks with a Covered Bonds Program (CBP) securitize ceteris paribus less of their assets.

Consistent with the argument of CBs being preferable as a funding tool, the difference is mainly

driven by banks characterized by a low level of assets liquidity. Proxies for a bank’s need to transfer

credit risk and/or meet regulatory capital requirements do not seem to be equally relevant. A

reduction in assets liquidity is associated with an expected increase in ABS issuance only for banks

without a CBP. This suggests that ABS are still regarded as a viable funding tool, but only for

banks not able to resort to CBs.

Our results thus show that the ability of issuing CBs is an important determinant of the decision

by European banks to securitize their assets. In the aftermath of the crisis, European institutions

such as the ECB and the newly created European Banking Association (EBA) have reckoned a

well-functioning ABS market as pivotal in providing Small and Medium Enterprises (SMEs) with

credit and banks with a diversified source of funding.11 A further reduction in the diversification of

funding sources could in fact result in higher bankruptcy costs (Colla, Ippolito, and Li, 2013) and

systemic risk (Oura, Gonzalez-Hermosillo, Chan-Lau, Gudmundsson, and Valckx, 2013). Moreover,

the current capital constraints associated to the use of ABS as a funding tool are passed-through

to financed companies (Carbo-Valverde, Degryse, and Rodrıguez-Fernandez, 2015). In this context,

several initiatives have been proposed to revitalize the market, and especially so for securities backed

by corporate loans (Aiyar, Al-Eyd, Barkbu, and Jobst, 2015). Despite this, some of the currently

proposed regulation seem to go in the direction of further favoring CBs over ABS. For example,

in the context of Basel III liquidity coverage requirements the EBA has proposed in 2013 to base

liquidity measurements on bid-ask spreads. This would once again favor CBs, even if alternative

11See for example the joint document by the BCE and the Bank of England: “The impaired EU securitization

market: causes, roadblocks and how to deal with them”. www.ecb.europa.eu/pub/pdf/other/ecb-boe_impaired_

eu_securitisation_marketen.pdf

15

measurements would suggest ABS are often as or even more liquid than CBs (Perraudin, 2014). In

pursuing their target to revitalize the European ABS market, regulators should keep in mind the

distorting effect of rules increasing the comparative advantage of CBs over ABS.

16

References

Acharya, Viral V, Philipp Schnabl, and Gustavo Suarez, 2013, Securitization without risk transfer,

Journal of Financial economics 107, 515–536.

Affinito, Massimiliano, and Edoardo Tagliaferri, 2010, Why do (or did?) banks securitize their

loans? evidence from Italy, Journal of Financial Stability 6, 189–202.

Aiyar, Mr Shekhar, Mr AJ Al-Eyd, Ms Bergljot Barkbu, and Andreas Jobst, 2015, Revitalizing Secu-

ritization for Small and Medium-Sized Enterprises in Europe . No. 15-17 (International Monetary

Fund Staff discussion note SDN/15/07).

Albertazzi, Ugo, Ginette Eramo, Leonardo Gambacorta, and Carmelo Salleo, 2015, Asymmetric

information in securitization: An empirical assessment, Journal of Monetary Economics 71, 33–

49.

Ambrose, Brent W, Michael LaCour-Little, and Anthony B Sanders, 2005, Does regulatory capi-

tal arbitrage, reputation, or asymmetric information drive securitization?, Journal of Financial

Services Research 28, 113–133.

An, Xudong, Yongheng Deng, and Stuart A Gabriel, 2011, Asymmetric information, adverse selec-

tion, and the pricing of cmbs, Journal of Financial Economics 100, 304–325.

Ashcraft, Adam B, and Til Schuermann, 2006, Understanding the securitization of subprime mort-

gage credit, Foundations and Trends R© in Finance 2, 191–309.

Ayotte, Kenneth, and Stav Gaon, 2011, Asset-backed securities: costs and benefits of “bankruptcy

remoteness”, Review of Financial Studies 24, 1299–1335.

Bannier, Christina E, and Dennis N Hansel, 2006, Determinants of banks’ engagement in loan

securitization, Discussion paper, Department of Finance, Goethe University Frankfurt am Main.

Basurto, Miguel A Segoviano, Bradley Jones, Peter Lindner, and Johannes Blankenheim, 2015,

Securitization: The Road Ahead (International Monetary Fund).

Bertaut, Carol, Laurie Pounder DeMarco, Steven Kamin, and Ralph Tryon, 2012, ABS inflows to

the united states and the global financial crisis, Journal of International Economics 88, 219 –

234 {NBER} Global.

Blum, Jurg M, 2008, Why basel II may need a leverage ratio restriction, Journal of Banking &

Finance 32, 1699–1707.

17

Calem, Paul S, and Michael LaCour-Little, 2004, Risk-based capital requirements for mortgage

loans, Journal of Banking & Finance 28, 647–672.

Calomiris, Charles W, and Joseph R Mason, 2004, Credit card securitization and regulatory arbi-

trage, Journal of Financial Services Research 26, 5–27.

Carbo-Valverde, Santiago, Hans Degryse, and Francisco Rodrıguez-Fernandez, 2015, The impact of

securitization on credit rationing: Empirical evidence, Journal of Financial Stability 20, 36–50.

Carbo-Valverde, Santiago, Francisco Rodrıguez Fernandez, and Richard Rosen, 2011, Are covered

bonds a substitute for mortgage-backed securities?, Federal Reserve Bank of Chicago WP 2011-14.

Cardone-Riportella, Clara, Reyes Samaniego-Medina, and Antonio Trujillo-Ponce, 2010, What

drives bank securitisation? the spanish experience, Journal of Banking & Finance 34, 2639–

2651.

Cebenoyan, A Sinan, and Philip E Strahan, 2004, Risk management, capital structure and lending

at banks, Journal of Banking & Finance 28, 19–43.

Celerier, Claire, and Boris Vallee, 2014, The motives for financial complexity: An empirical inves-

tigation, .

Colla, Paolo, Filippo Ippolito, and Kai Li, 2013, Debt specialization, Journal of Finance 68, 2117–

2141.

Coval, Joshua, Jakub Jurek, and Erik Stafford, 2009, The economics of structured finance, The

Journal of Economic Perspectives pp. 3–26.

Cragg, John G, 1971, Some statistical models for limited dependent variables with application to

the demand for durable goods, Econometrica 39, 829–44.

DeMarzo, Peter M., 2005, The pooling and tranching of securities: A model of informed intermedi-

ation, Review of Financial Studies 18, 1–35.

European Banking Authority, 2014a, EBA discussion paper on simple standard and transparent

securitisations, Discussion Paper, EBA/DP/2014/02.

, 2014b, Eba report on eu covered bond frameworks and capital treatment, Discussion paper,

.

European Central Bank, 2007, EU banking structures, October 2007.

, 2011, Recent developments in securitization, Discussion paper, .

18

European Commission, 2015, Proposal for a regulation of the european parlament and of the council,

directives 2009/65/EC, 2009/138/EC, 2011/61/EU and regulations (EU) No 1060/2009 and No

648/2012, Discussion paper, .

Fabozzi, Frank J, Henry A Davis, and Moorad Choudhry, 2007, Introduction to structured finance

. vol. 148 (John Wiley & Sons).

Farruggio, Christian, and Andre Uhde, 2015, Determinants of loan securitization in european bank-

ing, Journal of Banking & Finance 56, 12 – 27.

Gorton, Gary, and Andrew Metrick, 2012, Securitized banking and the run on repo, Journal of

Financial Economics 104, 425 – 451 Market Institutions, Financial Market Risks and Financial

Crisis.

Jiang, Wei, Ashlyn Aiko Nelson, and Edward Vytlacil, 2013, Securitization and loan performance:

Ex ante and ex post relations in the mortgage market, Review of Financial Studies p. hht073.

Jones, David, 2000, Emerging problems with the basel capital accord: Regulatory capital arbitrage

and related issues, Journal of Banking & Finance 24, 35–58.

Kempf, Alexander, Olaf Korn, and Marliese Uhrig-Homburg, 2012, The term structure of illiquidity

premia, Journal of Banking & Finance 36, 1381–1391.

Longstaff, Francis A., 2010, The subprime credit crisis and contagion in financial markets, Journal

of Financial Economics 97, 436 – 450.

Maddaloni, Angela, and Jose-Luis Peydro, 2011, Bank risk-taking, securitization, supervision, and

low interest rates: Evidence from the euro-area and the us lending standards, Review of Financial

Studies 24, 2121–2165.

Norddeutsche Landesbank, 2013, Covered bond issuer guide, .

Oura, Hiroko, Brenda Gonzalez-Hermosillo, Jorge A Chan-Lau, Tryggvi Gudmundsson, and Nico

Valckx, 2013, Changes in bank funding patterns and financial stability risks, IMF Global Financial

Stability Report.

Packer, Frank, Ryan Stever, and Christian Upper, 2007, The covered bond market, BIS Quarterly

Review, September.

Pagano, Marco, and Paolo Volpin, 2012, Securitization, transparency, and liquidity, Review of

Financial Studies 25, 2417–2453.

Perraudin, William, 2014, Covered bond versus abs liquidity: A comment on the EBAs proposed

hqla definition, Discussion paper, .

19

Van Rixtel, Adrian ARJM, and Gabriele Gasperini, 2013, Financial crises and bank funding: recent

experience in the euro area, .

Volk, B, and F Will, 2012, Covered bonds versus senior unsecured bank debt and rmbs, European

Covered Bond Council: European Covered Bond Fact Book, pp. 206–222.

Wang, Yihui, and Han Xia, 2014, Do lenders still monitor when they can securitize loans?, Review

of Financial Studies p. hhu006.

Wooldridge, Jeffrey M, 2010, Econometric analysis of cross section and panel data (MIT press).

20

Figure 1: Average marginal effects

This figure presents the average marginal effects on the unconditional expected value of SEC ratio. Each of the three

rows focuses on a different bank characteristic x, where x is Liquidity ratio (first row), Loss to loans (second row),

or Tier 1 ratio (third row). Effects presented in the first, second, and third row are based on coefficient estimates

for Model (i), (ii), and (iii) respectively as presented in Table 6. The column on the left presents the (average)

difference in the unconditional expected level of securitization between banks with and without a covered bond

program, computed for different levels of x. Negative values indicate a lower securitization activity by banks with a

covered bond program. The column on the right presents the difference in the average marginal effect of x between

banks with and without a covered bond program, also computed for different levels of x. Positive values indicate a

higher (i.e., smaller negative) marginal effect for banks with a covered bond program. Areas in grey represent the

95% confidence interval.

-1.5

-1-.

50

.5

0 20 40 60 80 100Liquidity ratio

-.6

-.4

-.2

0.2

.4

0 10 20 30 40Loss to loans

-10

-50

5

0 10 20 30 40 50Tier 1 ratio

Average marginal effect of CBP

0.0

2.0

4.0

6

0 20 40 60 80 100Liquidity ratio

0.0

5.1

.15

.2.2

5

0 10 20 30 40Loss to Loans

-.6

-.4

-.2

0.2

0 10 20 30 40 50Tier 1 ratio

Difference in average marginal effect of x, CBP vs. no CBP

21

Figure 2: Average marginal effect of liquidity: banks with and without a covered bond program

This figure presents the average marginal effect of Liquidityratio on the unconditional expected value of SECratio for

banks with (dotted line) and without (continuous line) a covered bond program, for different levels of Liquidityratio.

Marginal effects are computed based on coefficient estimates for Model (i) as presented in Table 6. Areas in grey

represent the 95% confidence interval.-.

06-.

04-.

020

.02

0 20 40 60 80 100Liquidity ratio

CBP=0 CBP=1

22

Table 1: Main differences between Asset-Backed Securities and Covered Bonds

This table summarizes the main differences between Asset-Backed Securities and Covered Bonds, as discussed in

Section 2.2.

Asset-Backed Securities Covered Bonds

Issuer Special Purpose Vehicle (SPV) Bank originating the assets

Investors claim Over assets owned by the SPV and the

associated cash flows

Over collateral and the issuer

Eligible assets No or little restriction on eligible assets Defined by national laws

Structure Customized Standardized. Face value typically

repaid at maturity

Balace Sheet treatment Assets are sold to SPV. Credit Risk is

transferred

Pledged cover assets remain in the

balance sheet of the issuer

Availability No or little restriction Covered Bond Programs subjected to

national regulators approval

Collateral pool Typically static/revolving Dynamic. Quality of the collateral has

to be preserved

Government guarantee None Implicit in support for the issuer

23

Table 2: Sample distribution by year and country

This table presents the distribution of bank-year observations in the dataset by country (rows) and year (columns).

Country\Year 2007 2008 2009 2010 2011 2012 2013

Austria 10 13 15 15 16 16 16

Belgium 6 6 8 8 8 7 5

Germany 11 14 18 19 20 22 20

Spain 21 22 25 25 24 23 15

Finland 2 2 2 1 4 4 4

France 10 10 8 5 5 7 7

Greece 5 6 7 7 6 7 6

Ireland 6 6 6 6 6 6 6

Italy 9 13 13 13 13 12 13

Netherlands 8 8 8 9 10 10 8

Portugal 7 6 6 6 6 6 6

24

Table 3: Characteristics of securitization activity for banks in the sample

This table presents descriptive statistics on securitization activities for bank-year observations in the sample. Bank-

year observations where no issuance of structured finance products (either to be retained or distributed) is observed

are excluded. Underlying asset pool refers to the value-weigthed share of securities backed by different types of assets

for each bank-year, expressed in percentage points. Total amounts (in Euro billions) statistics are presented for all

Asset-Backed Securities, excluding securities created to be retained.

N Mean Std. Dev. 5% perc. Median 95% perc.

No. Deals 74 6.74 18.69 1 2 20

No. Tranches 74 17.15 26.66 3 10 57

Total amount (bln. Euro) 74 2.39 3.98 0.06 1.10 13.00

Underlying asset pool (%)

Auto 74 15.37 33.06 0.00 0.00 100.00

Collateralized Debt Obligations 74 11.46 28.68 0.00 0.00 100.00

Commercial Mortgage-Backed Securities 74 0.96 5.64 0.00 0.00 0.00

Credit Cards receivables 74 0.00 0.00 0.00 0.00 0.00

Consumer Loans 74 6.19 18.15 0.00 0.00 63.64

Leases 74 0.88 4.75 0.00 0.00 0.60

Residential Mortgage-Backed Securities 74 49.95 44.85 0.00 52.68 100.00

SMEs Collateralized Loan Obligations 74 13.97 29.13 0.00 0.00 100.00

Other 74 1.23 4.7 0.00 0.00 9.02

25

Table 4: Descriptive statistics

This table presents descriptive statistics for the variables included in this study. SEC ratio is the ratio of Euro-

denominated structured finance products (excluding retained securities) over bank’s total assets, expressed in per-

centage points. SEC is an indicator equal to 1 when SEC ratio > 0 and 0 otherwise. CBP is an indicator equal to 1

for banks with a covered bonds program and 0 otherwise. Liquidity ratio is computed as 1 minus the ratio between

net loans and total assets. Loss to loans is the ratio of loan loss reserves to gross loans. Tier 1 ratio is the ratio of

accounting value of bank’s equity over its risk-weighted assets. ROE is the ratio of bank’s operating profit to the

accounting value of equity. Liquidity ratio, Loss to loans, Tier 1 ratio, and ROE are all expressed in percentage

points and lagged 1 year. LNTA is the natural logarithm of the (1-year lagged) bank’s total assets, expressed in

Euro millions. Commercial is an indicator equal to 1 if the bank is a commercial bank and 0 otherwise. Slope is

the difference between 10-year and 3-year government bond yields. GDP is the country GDP growth rate.

Variable N Mean Std. Dev. 5% perc. Median 95% perc.

SEC 775 0.10 0.29 0 0 1

SEC ratio 775 0.18 0.91 0.00 0.00 0.65

CBP 775 0.55 0.50 0 1 1

Liquidity ratio 775 42.54 19.63 16.86 38.31 80.65

Loss to loans 775 2.72 3.01 0.22 2.16 6.83

Tier 1 ratio 775 10.92 5.85 6.31 9.67 18.2

ROE 775 0.03 2.84 -0.39 0.08 0.23

LNTA 775 10.83 1.71 8.04 10.79 13.73

Commercial 775 0.68 0.47 0 1 1

Slope 775 2.42 2.70 0.14 2.16 3.75

GDP 775 0.01 0.03 -0.05 0.02 0.06

26

Table 5: Covered bond program and securitization activity

This table presents coefficient estimates of a Cragg (i) and a Tobit II (ii) model for the securitization activity of the

bank. The dependent variables are SEC and SEC ratio for the selection (columns (1) and (4)) and the amount

(columns (2) and (5)) equation of the two models respectively. SECt−1 is the 1-year lag of SEC. All of the other

variables are as defined in Table 4. Standard errors robust to heteroskedasticity and clustered by bank are reported

in parenthesis. No. observations is the number of bank-year observations included. No. banks is the number of

distinct banks. No. issuing is the number of bank-year observations where SEC = 1. Columns (3) and (6) report

the Average Marginal Effect (AME) of each variable on the unconditional expected value of SEC ratio for Model

(i) and (ii) respectively. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% confidence level

respectively.

(i) Cragg model (ii) Tobit II model

(1) Selection (2) Amount (3) AME (4) Selection (5) Amount (6) AME

CBP -0.520** -0.603 -0.130** -0.292 -0.601 -0.110*

(0.259) (0.841) (0.062) (0.200) (0.484) (0.056)

Liquidity ratio -0.030*** -0.073** -0.009*** -0.024*** -0.011 -0.005***

(0.010) (0.032) (0.003) (0.006) (0.016) (0.002)

Loss to loans -0.184*** -1.018** -0.072** -0.111** -0.136 -0.031*

(0.067) (0.471) (0.030) (0.051) (0.170) (0.017)

Tier 1 ratio 0.007 -0.132 -0.003 0.021 -0.106 -0.007

(0.027) (0.205) (0.012) (0.019) (0.150) (0.013)

ROE -0.013 4.162 0.137 0.013 3.087 0.296

(0.017) (4.683) (0.325) (0.059) (2.284) (0.212)

LNTA 0.607*** -1.825*** 0.064* 0.432*** -0.916*** -0.016

(0.140) (0.558) (0.034) (0.100) (0.286) (0.030)

Commercial 0.374 -0.276 0.065 0.288 0.554 0.092

(0.316) (0.790) (0.058) (0.242) (0.596) (0.057)

Slope 0.136 1.810* 0.089* 0.128 -0.049 0.017

(0.107) (0.998) (0.053) (0.095) (0.203) (0.030)

GDP 1.482 5.969 0.506 -0.293 3.957 0.328

(5.262) (45.566) (1.846) (5.088) (14.198) (1.793)

SECt−1 1.821***

(0.271)

Year indicators Yes Yes Yes Yes

σ 1.613 1.500

λ -0.195

No. Observations 775 775

No. Banks 134 134

No. Issuing 74 74

27

Table 6: Covered bond program and securitization determinants

This table presents coefficient estimates of a Tobit II models for the securitization activity of each bank. The

dependent variables are SEC and SEC ratio for the selection (columns (1), (3), and (5)) and the amount (columns

(2), (4), and (6)) equation of the model respectively. Interaction represents for Models (i), (ii), and (iii) respectively

the cross-product of CBP and Liquidity ratio, Loss to loans, or Tier 1 ratio. SECt.1 is the 1-year lag of SEC.

Other controls include ROE, LNTA, Slope, and GDP . All other variables are as defined in Table 4. Standard errors

robust to heteroskedasticity and clustered by bank are reported in parenthesis. No. observations is the number of

bank-year observations included. No. banks is the number of distinct banks. No. issuing is the number of bank-year

observations where SEC = 1. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% confidence level

respectively.

(i) Liquidity ratio (ii) Loss to loans (iii) Tier 1 ratio

(1) Selection (2) Amount (3) Selection (4) Amount (5) Selection (6) Amount

CBP -0.818** -3.030*** -0.530 -1.522** -1.441*** 0.357

(0.389) (0.882) (0.355) (0.731) (0.545) (1.773)

Interaction 0.013* 0.057*** 0.116 0.465* 0.108** -0.099

(0.007) (0.015) (0.102) (0.275) (0.054) (0.184)

Liquidity ratio -0.030*** -0.030** -0.023*** -0.011 -0.027*** -0.008

(0.008) (0.015) (0.006) (0.016) (0.006) (0.016)

Loss to loans -0.107** -0.044 -0.185** -0.340 -0.113** -0.140

(0.052) (0.153) (0.087) (0.238) (0.053) (0.167)

Tier 1 ratio 0.020 -0.097 0.019 -0.041 -0.042 -0.034

(0.021) (0.125) (0.020) (0.151) (0.045) (0.248)

SECt−1 1.818*** 1.821*** 1.778***

(0.273) (0.273) (0.255)

Other controls Yes Yes Yes Yes Yes Yes

Year indicators Yes Yes Yes Yes Yes Yes

σ 1.351 1.426 1.446

λ -0.344 -0.215 -0.188

No. Observations 775 775 775

No. Banks 134 134 134

No. Issuing 74 74 74

28